Few-shot Semantic Segmentation with Support-induced Graph Convolutional Network

TL;DR

This paper introduces SiGCN, a novel graph convolutional network for few-shot semantic segmentation that captures rich context and handles appearance variations, achieving state-of-the-art results on standard benchmarks.

Contribution

The paper proposes a Support-induced Graph Convolutional Network with modules for explicit context reasoning and instance association, improving robustness in few-shot segmentation.

Findings

Achieves state-of-the-art performance on PASCAL-5i and COCO-20i datasets.

Effectively captures multi-level query object parts and high-order instance context.

Enhances robustness to appearance variations in few-shot segmentation.

Abstract

Few-shot semantic segmentation (FSS) aims to achieve novel objects segmentation with only a few annotated samples and has made great progress recently. Most of the existing FSS models focus on the feature matching between support and query to tackle FSS. However, the appearance variations between objects from the same category could be extremely large, leading to unreliable feature matching and query mask prediction. To this end, we propose a Support-induced Graph Convolutional Network (SiGCN) to explicitly excavate latent context structure in query images. Specifically, we propose a Support-induced Graph Reasoning (SiGR) module to capture salient query object parts at different semantic levels with a Support-induced GCN. Furthermore, an instance association (IA) module is designed to capture high-order instance context from both support and query instances. By integrating the proposed two modules, SiGCN can learn rich query context representation, and thus being more robust to appearance variations. Extensive experiments on PASCAL-5i and COCO-20i demonstrate that our SiGCN achieves state-of-the-art performance.